BACKGROUND AND SUMMARY OF THE INVENTION
[0001] The present invention relates generally to the initial synchronization methods and
more specifically to an improvement in the steps of the initial synchronization methods.
[0002] Although the method will be described for mobile WiMAX systems (IEEE 802.16e, WiBro)
as an example, it is applicable to other communication systems and protocols. The
initial synchronization methods generally include a frame boundary search, fractional/integer
frequency offset estimation, and segment/cell ID searches.
[0003] The present method of initial synchronization of a communication signal includes
the steps of symbol boundary search, fractional frequency offset estimation, fractional
frequency offset compensation, frame boundary search, integer frequency offset estimation,
integer frequency offset compensation, preamble segment ID search and preamble cell
ID search.
[0004] The symbol boundary search includes estimating the boundary of a present data symbol
by a correlation index for the present data symbol and the correlation index for the
next data symbol. The combined correlation index is
where i denotes the correlation index,
G the cyclic prefix length,
y the observed time domain samples, and
NFFT the size of the symbol. The combined correlation index
i is calculated iteratively as follows:
[0005] where
and
[0006]
[0007] The frame boundary search includes identifying the preamble symbol in the symbols
found in the symbol boundary search to determine the frame boundary. Identifying the
preamble symbol includes grouping the subcarriers into K subgroups of N consecutive
subcarriers, where K is the number subcarries that define a specific segment group
of subcarriers; collecting the distributed energies on subcarriers; and making a decision
if the current symbol is preamble based on a threshold that is estimated by stochastic
count process model.
[0008] The integer frequency offset is estimated from the pilot subcarriers of the frame
control header of the frame without decoding the down load MAP.
[0009] The preamble segment ID search is based on:
where Pr
emableCarrier , the group index of
N groups
n =0,1,2..
N-1, and the subcarrier index of a K length PN sequence
k={0,1,2...
K-1}.
[0010] The preamble cell ID search includes estimating the symbol timing offset
by:
[0011]
where the group index of
N groups
n =0,1,2..
N-1, cell ID of
R cell IDs in a segment group
r ∈ {0,1,..,
R-1},
mk and
mk+1 are two neighboring subcarrier positions, the subcarrier index of a K length PN sequence
k={0,1,2..
K-1} and
represents the modulated PN sample for the preamble.
[0012] The preamble cell ID is estimated by:
where
m ∈ {
non-zero subcarriers},
NFTT is the symbol size.
[0013] These and other aspects of the present invention will become apparent from the following
detailed description of the invention, when considered in conjunction with accompanying
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 is a flow chart of method of initial synchronization of a communication
signal according to the present disclosure.
[0015] Figure 2 is a diagram of an example of a TDD frame of the prior art.
[0016] Figure 3 is a diagram of an example of a subcarrier index of the prior art.
[0017] Figure 4 is a diagram of a double correlation of the estimate the symbol boundary
according to the present disclosure.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0018] The method of initial synchronization of a communication signal is illustrated in
Figure 1. The initial synchronization process begins at Step 10 to perform the symbol
boundary search and fractional frequency offset estimation using a double correlation.
Next the fractional frequency offset is compensated at Step 12. The frame boundary
search is performed by preamble identification at Step 14. If a preamble has not been
found at Step 16, the process goes back to the Step 10 via 18 to again do a double
correlation.
[0019] If a preamble has been found at Step 16, the integer frequency offset is estimated
using the frame control head or FCH at Step 20. Next the integer frequency offset
compensation for the preamble is performed at Step 22. Finally, a segment/cell ID
search is performed on the preamble at Step 24. This is the end of the synchronization
process.
[0020] Each frame in the downlink transmission begins with a preamble followed by down load
DL and up load UL transmission periods as shown in Figure 2. In the frame, the TTG
shall be inserted between the downlink and uplink, and RTG at the end of each frame
to allow the BS to turn around. In IEEE 802.16e WMAN-OFDMA, the duration of one physical
frame ranges from 2 ms to 20 ms, and the FFT sizes of an OFDMA symbol are defined
as 128, 512, 1024, and 2048. See IEEE 802.16e - 2005, Standard for local and metropolitan
area networks Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access
Systems. While in WiBro, the duration of one physical and the FFT size of OFDMA symbol
are fixed as 5 m and 1024, respectively. See
Specifications for 2.3 GHz band Portable Internet Service (TTAS. KO-06.0082/
R1)..
[0021] Although the initial synchronization algorithms is described for the WiBro system
as an example, all these algorithms can be applied to 802.16e WMAN-OFDMA as well.
[0022] The first symbol of DL burst in every frame is the preamble. In the preamble, there
are N or in this example three different segment groups, and each group has different
PN sequences. The indices of these groups can be defined as
[0023] where the group index of
N groups
n=0,1,2..
N-1, and the subcarrier index of a K length PN sequence
k= {0, 1,2 ...
K-1} which will be mapped through.
[0024] Figure 3 shows the carrier indexing for the case of
n = 0. Data in every third position is separated by two null positions.
[0025] For example, in 1024 FFT mode, there are 114 different PN sequences (see Table 1),
and out of them each 38 sequences are defined for a specific subcarrier group. Within
a subcarrier group, each of the 38 sequences indicates a specific cell. There are
284 subcarriers defined for a specific segment group. The index variable
n and
k in equation (1) determines the segment of the cell and subcarrier index of the preamble
subcarriers, respectively.
Table 1. PN sequence per segment & cell ID for 1024 FFT mode.
Index |
IDcell |
Segment |
Series to modulate (im hexadecimal format) |
0 |
0 |
0 |
A6F294537B285E1844677D133E4D53CCB1F182DE00489E53E6B6E77 065C7EEE7D0ADBEAF |
1 |
1 |
0 |
668321CBBE7F462E6C2A07E8BBDA2C7F7946D5F69E35AC8ACF7D6 4AB4A33C467001F3B2 |
2 |
2 |
0 |
1C75D30B2DF72CEC9117A0BD8EAF8E0502461FC07456AC906ADE0 3E9B5AH5E1D3F92C6E |
3 |
3 |
0 |
5F9A2E5CA7CC69A5227104FB1CC2262809F3B10D0542B9BDFDA4A 73A7046096DF0E8D3D |
4 |
4 |
0 |
82F8A0AB918138D84BB86224F6C342D81BC8BFE791CA9EB5409615 9D672E91C6E1032F |
5 |
5 |
0 |
EE27E39B84CCF15BB1365EF90D478CD2C49EE8A70DE368EED7C94 20BOC6FFAF9AF035FC |
|
|
|
|
110 |
14 |
2 |
29E74579472FDD8FFC2700B2BF33C649989DD8153093A7CA08B50F7 A5E4BAED108A0F0D |
111 |
15 |
0 |
A27F29D8D6CCD7EB4BBEM303E9E95802DB98BFD5B8ED03B8830 4359D923EC108CA3C8 |
112 |
16 |
1 |
3FE70E26FA00327FE3B2BE6BC5D5014F588F09C17D222C146DD68B 4824692A65188C76 |
113 |
17 |
2 |
41E91307EC58801CFF2C7E9CFEFBEB71681FAE2BEAEC72D4E4556 E99345D3BA4B369B59 |
[0026] The PN sequence to be assigned onto the preamble carrier set is modulated as
where
wk denotes the PN sequence in Table 1. Finally, the modulated preamble sequence is assigned
to preamble subcarriers according to the index defined in equation (1).
[0027] In addition to modulated preamble subcarriers, for 1024 FFT mode, there will be 86
guard band subcarriers on the left and right.
[0028] The frame boundary search consists of two steps; the symbol boundary search and preamble
identification. Once stream of filtered samples are available, the symbol boundary
search block compute correlations between one sample to other sample with distance
of symbol size. This correlation utilizes the repeating characteristic of the guard
period (CP: Cyclic Prefix). Let
NFFT and
G be the OFDMA symbol size and the CP size. The correlation window size is set to
NFFT +
G and within the window, a position of the correlated samples which provides the maximum
correlator outputs is regarded as the symbol boundary. As discussed below, the frame
boundary search scheme based on the symbol boundary search result.
[0029] The symbol boundary search relies on the characteristic of the guard period repetition
in OFDM symbol. In this case, the ML estimate of the Symbol boundary is well known
and is given by:
where
i, G,
y, and
NFFT denote the correlation index, the cyclic prefix length, the observed time domain
samples, and the number of the OFDM symbol subcarriers, respectively. The correlation
of the repeating pattern of guard period for the next data symbol and add this additional
correlation result to equation (3). This is the double correlator shown in Figures
1 and 4.
[0030] The estimation based on the double correlator can be shown as
[0031] Simply speaking, it is doubling the amount of statistics to provide better ML estimation
results. The correlation in equation (4) can be calculated in iterative manner like
below:
where
and
[0032] The number of computations in equation (4) does not require twice the number of computations
than in equation (3). To explain the computational complexity for equation (4), let
be the correlator outputs through sample
n and n +
NFFT within a correlation window
t, where
n =0, 1, ..,
NFFT +
G -1 and t = 0, 1, 2, 3,.... Also, let
When the correlations,
St, 0 <
t, are calculated for a group of samples covered in
t-th correlation window, the double correlation in equation (4) can be computed by
adding the previous correlation outputs,
St-1 to the current correlation outputs for
St. Thus, the computations in equation (4) only require extra
NFFT +
G additions of the computations in equation (3).
[0033] That is, the terms
and
in equation (5) are already computed at previous correlator at window
t-1 and can be added to the current correlator output to make double correlation.
[0034] The difference in the slot boundary estimation performance between the method in
equation (3) and equation (4) is significant. Following Table 2 shows the variances
for the double and single correlator for the case of SNRs; 3 dB, 6 dB, and 9 dB. The
number of trials were 10,000 per each SNR case.
Table 2
|
3dB |
6dB |
9dB |
Variance of the Symbol boundary estimator (Single Correlator) |
1.30 |
0.86 |
0.71 |
Variance of the Symbol boundary estimator (Double Correlator) |
0.60 |
0.40 |
0.32 |
[0035] The results clearly show that the variances of the symbol boundary estimator from
double correlator are about two times smaller than one from single correlator.
[0036] In the following table 3, another measurement is presented in which the performances
of double and single correlator can be compared. The metric is defined as
[0037] In Table 3, this metric is computed 10, 000 times for SNR of 3 dB, 6 dB, and 9 dB.
Table 3
|
3dB |
6dB |
9dB |
Rate_Miss (Single Correlator) |
0.42 |
0.36 |
0.33 |
Rate_Miss (Double Correlator) |
0.28 |
0.22 |
0.19 |
[0038] Once the symbol boundary is estimated, the frame boundary has to be determined. In
the system, since the first symbol in the frame is the preamble, searching for the
frame boundary is the same as identifying the preamble symbol out of the symbols whose
boundary has been found.
[0039] Without loss of generality, the search procedure based on the WiBro system (1024
FFT mode) will be explained. In the WiBro system, after disregarding guard bands,
a preamble is composed of 852 subcarriers (284 non-zero subcarriers + 568 zero subcarriers).
There are 284 subcarriers defined for a specific segment group,
j, j = 0,1,2. Let us define a subgroup which includes 3 consecutive subcarriers. There will
be 284 subgroups and within each subgroup, the index
j, j = 0,1,2, represents corresponding segment ID. The approach for identifying preamble
is based on a count process, in which the energies per each subcarrier are computed
and the computed energies of three consecutive subcarriers are grouped as three vectors,
say
as discussed above. Within each group, a subcarrier position,
j, j = 0,1,2, corresponds to the maximum energy in the group are searched and counted.
[0040] For clarification purpose, this procedure is explained in the following example.
Suppose the sequence of subcarrier energies for the FFTed and guard band removed preamble
symbol is (1.0, 3.6, 0.9, 1.2, 3.3, 0.7, ......,2.1, 1.3, 0.7, 1.1, 3.3, 0.7). First,
the sequence of subcarrier energies are grouped as three vectors in consecutive manner
and let
j̃k denote the position of subcarrier which provide maximum energy in group
k. The following is the count process
[0041] After counting
j̃k, a dominant
j,
j = 0,1,2, in the groups is determined from the number of its occurrences. The number
of occurrences for
j̃k ∈ {0,1,2} can be interpreted as the weight of its significance, and the normalized
version of these weights can be used as the probabilities of
j, j = 0,1,2, being a dominating subcarrier position. These probabilities can be well
modeled by a conjugate prior distribution to multinomial distribution, Dirichlet distribution
described in
J. M. Bernardo and A. F. M. Smith, "Bayesian Theory", Wiley 1994.
[0042] A continuous random vector
x=(
x1,
x2,...,
xk) has a Dirichlet distribution of dimension k, with parameters α
=(
a1,
a2,...,
ak+1) (α
i > 0,
i = 1,...,
k =1) if its probability density
Dk(
x|α), 0 <
xi < 1 and
x1 +
x2 + ... +
xk < 1, is
where c is the normalization constant and is defined by
[0043] The mean vector are given by
[0044] Let
Pj,j = 0,1,2, be the probability of
j, j = 0,1,2, being dominant in groups, and let
Nj, j = 0,1,2, be the counted numbers which says how many times
j being dominant in 284 groups. Then, the Dirchlet distribution of
Pj, j = 0,1,2, can be expressed by
[0045] The mean of the distribution is used for the probability estimation. That is,
[0046] The estimated probability
P̂j, j = 0,1,2, is used as the threshold for determining the preamble. Following Table 4
shows the estimated probability of
j, j = 0,1,2, for SNR of 0 dB, 3 dB, and 6 dB when the true segment ID is 1. The number
of trials were 10,000 for each SNR case.
Table 4
|
0 dB |
3 dB |
6 dB |
P̂0 |
0.02 |
0.0018 |
0.0 |
P̂1 |
0.96 |
0.9916 |
1.0 |
P̂2 |
0.02 |
0.0018 |
0.0 |
[0047] The results in the table above shows us that even in hostile environment (SNR = 0
dB case), the estimated probability of dominating subcarrier position is as clear
as 0.96.
[0048] The probability can be used as the threshold for preamble decision. For example,
the probability 0.96 can be interpreted as like one subcarrier position occurred in
about 272 times in SNR 0 dB. Thus, in SNR 0 dB environment, if the number of occurrences
of a subcarrier position is over 272 times, the current symbol is the preamble.
[0049] The frequency offset is estimated after symbol synchronization. In the offset, there
are integer parts and fractional parts, and both have to be estimated. Let ε,
εf, and ε
l denote frequency offset, integer frequency offset, and fractional frequency offset,
respectively.
[0050] The relationship among frequency offset, the integer frequency offset, and fractional
frequency offset can be given as
[0051] ML estimation of the fractional frequency offset based on correlator outputs have
been well known [2]. The fractional frequency offset ε
f can be estimated from
where
C(
î) is the maximum correlation calculated in (5), D denotes the delay, and
Ts the sample time.
[0052] The pilot subcarriers in FCH (Frame Control Header) is used to estimate the integer
frequency offset. Since the subchannelization scheme used for FCH is always PUSC,
we do not need to decode DL MAP to check the subchannelization mode of FCH. Thus,
upon reception of the symbol following the preamble, the exact location of pilot subcarriers
is after taking FFT of the symbol. In the WiBro system, the power of each pilot tone
should be boosted 2.5 dB higher than the average power level of data tones. With this
constraint, the integer frequency offset which can cause cyclic shift of subcarrier
position in a symbol is estimated.
[0053] The segment ID and cell ID search are performed after preamble identification. The
post-FFT processing for both segment ID and the cell ID search are used. As shown
in Figure 1, once the integer frequency offset is compensated for an identified preamble
symbol, segment ID search is performed on the FFTed preamble symbol. After the segment
ID is estimated, the number of cell ID candidates can be reduced by factor 3. That
is, cell ID search can be performed within a decided segment group.
[0054] Let
yFFT be the FFTed frequency samples of the preamble after removing guard subcarriers.
The segment ID can be then searched based on the following criterion (See equation
(1) for the description of index variables):
[0055] After the segment ID is determined, the number of candidates for the true preamble
sequence is reduced to one third of the number. With regard to each candidate sequence,
the symbol timing offset is estimated from the possible frame boundary estimation
error. Symbol timing offset can be estimated by averaging the phase differences among
neighbor frequency samples:
where
denotes the symbol timing error estimate based on assumption that the preamble is
member of a segment group
n, n=0,1,2, and cell ID,
r ∈ {0,1,..,
R-1),
R denotes the number of cell IDs in a segment group. The
mk and
mk+1,
k = 0,1,..,
K-1 (see equation (1) for the description of
K), are two neighboring subcarrier positions, and
represents the modulated PN sample for the preamble. See K. Nikitopoulos and A. Polydoros,
"Post-FFT Fine Frame Synchronization for OFDM system" in
VTC 1997.
[0056] The average operation in equation (9) can be replaced by simple summation. Also since
the phase difference is only of interest, calculation can be further simplified as:
[0057] For each
r, the symbol timing error can be compensated (or corrected) by
where
m ∈ {
non -
zero subcarriers},
r = {0,1,..,
R-1}.
[0058] The cell ID is estimated by choosing the maximum output of the dot products (cross-correlating
equation (7) and
r modulated pilot patterns with lag zero). That is,
[0059] Although the present invention has been described and illustrated in detail, it is
to be clearly understood that this is done by way of illustration and example only
and is not to be taken by way of limitation. The scope of the present invention is
to be limited only by the terms of the appended claims.
1. A method of initial synchronization of a communication signal including the steps
of symbol boundary search, fractional frequency offset estimation, fractional frequency
offset compensation, frame boundary search, integer frequency offset estimation, integer
frequency offset compensation, preamble segment ID search and preamble cell ID search,
wherein the symbol boundary search comprises:
estimating the boundary of a present data symbol by a correlation index for the present
data symbol and the correlation index for the next data symbol.
2. The method of Claim 1, wherein the combined correlation index is
where i denotes the correlation index,
G the cyclic prefix length,
y the observed time domain samples, and
NFFT the size of the symbol.
3. The method of Claim 2, wherein the combined correlation index i is calculated iteratively
as follows:
where
and
4. The method of Claim 1, wherein the frame boundary search includes identifying the
preamble symbol in the symbols found in the symbol boundary search to determine the
frame boundary.
5. The method of Claim 4, wherein identifying the preamble symbol includes grouping the
subcarriers into K subgroups of N consecutive subcarriers, where K is the number subcarries
that define a specific segment group of subcarriers; collecting the distributed energies
on subcarriers; and making a decision if the current symbol is preamble based on a
threshold that is estimated by stochastic count process model.
6. The method of Claim 1, wherein the integer frequency offset estimation is derived
from the pilot subcarriers of the frame control header of the frame after locating
the preamble symbol.
7. The method of Claim 1, wherein the integer frequency offset estimation is derived
from the pilot subcarriers of the frame control header of the frame without decoding
the down load MAP.
8. The method of Claim 1, wherein the preamble segment ID search and the preamble cell
ID search are performed after the integer frequency offset compensation of the identified
preamble.
9. The method of Claim 8, wherein the preamble segment ID search is performed before
the preamble cell ID search.
10. The method of Claim 1, wherein the preamble segment ID search is based on:
where
the group index of
N groups
n=0,1,2..
N-1, and the subcarrier index of a K length PN sequence
k= {0,1,2...
K-1}.
11. The method of Claim 1, wherein the preamble cell ID search includes estimating the
symbol timing offset
by:
where the group index of
N groups
n=0, 1,2..
N-1, cell ID of
R cell IDs in a segment group
r ∈ {0,1,..,
R-1},
mk and
mk+1 are two neighboring subcarrier positions, the subcarrier index of a K length PN sequence
k={0,1,2...
K-1} and
represents the modulated PN sample for the preamble.
12. The method of Claim 11, wherein the preamble cell ID is estimated by:
where
m ∈ {
non-zero subcarriers},
NFTT is the symbol size.
13. A method of initial synchronization of a communication signal including the steps
of symbol boundary search, fractional frequency offset estimation, fractional frequency
offset compensation, frame boundary search, integer frequency offset estimation, integer
frequency offset compensation, preamble segment ID search and preamble cell ID search,
wherein the frame boundary search includes identifying the preamble symbol in the
symbols found in the symbol boundary search to determine the frame boundary, and identifying
the preamble symbol comprises:
grouping the subcarriers into K subgroups of N consecutive subcarriers, where K is
the number subcarries that define a specific segment group of subcarriers;
collecting the distributed energies on subcarriers; and
making a decision if the current symbol is preamble based on a threshold that is estimated
by stochastic count process model.
14. A method of initial synchronization of a communication signal including the steps
of symbol boundary search, fractional frequency offset estimation, fractional frequency
offset compensation, frame boundary search, integer frequency offset estimation, integer
frequency offset compensation, preamble segment ID search and preamble cell ID search,
wherein the integer frequency offset estimation is derived from the pilot subcarriers
of the frame control header of the frame after locating the preamble symbol.
15. The method of Claim 14, wherein the integer frequency offset estimation is derived
from the pilot subcarriers of the frame control header of the frame without decoding
the down load MAP.
16. A method of initial synchronization of a communication signal including the steps
of symbol boundary search, fractional frequency offset estimation, fractional frequency
offset compensation, frame boundary search, integer frequency offset estimation, integer
frequency offset compensation, preamble segment ID search and preamble cell ID search,
wherein the preamble segment ID search is based on:
where
the group index of
N groups
n=0,1,2.
N-1, and the subcarrier index of a K length PN sequence
k={0,1,2...
K -1}.
17. A method of initial synchronization of a communication signal including the steps
of symbol boundary search, fractional frequency offset estimation, fractional frequency
offset compensation, frame boundary search, integer frequency offset estimation, integer
frequency offset compensation, preamble segment ID search and preamble cell ID search,
wherein the preamble cell ID search includes estimating the symbol timing offset
by:
where the group index of
N groups
n=0,1,2..
N-1, cell ID of
R cell IDs in a segment group
r ∈ {0,1,..,
R-1},
mk and
mk+1 are two neighboring subcarrier positions, the subcarrier index of a K length PN sequence
k={0,1,2...
K-1} and
represents the modulated PN sample for the preamble.
18. The method of Claim 17, wherein the preamble cell ID is estimated by:
where
m ∈ {
non-zero subcarriers},
NFTT is the symbol size.